Agent integrations (MCP)

Endpoint

The endpoint speaks MCP over Streamable HTTP. POST requests must sendAccept: application/json, text/event-stream(compatible MCP clients do this automatically).

OAuth setup

Login redirects (next= behavior)

Available tools

{ profile: { id, full_name, goal, daily_calorie_target, food_preference, ... } }

{ plan: { id, title, daily_calories, plan_data: DietPlan, created_at } } | text: 'No diet plan generated yet.'

{ plan: DietPlan, plan_id: uuid, reused: boolean } — DietPlan = { title, daily_calories, macros, bmi, bmi_category, tips[], health_highlights[], days[{ day, meals: {breakfast,lunch,snack,dinner} }] }

{ meals: MealLog[] }

{ meal: MealLog }

{ entry: WeightLog }

{ entry: WaterLog }

{ subscription: { plan, status, current_period_end } | null }

Example: generate_diet_plan over MCP

A full round-trip from an MCP client (ChatGPT / Claude / Cursor) callinggenerate_diet_plan, followed by the structured response the tool returns. This is exactly the shape you'll receive — use it to preview how highlights and the daily meal schedule are formatted.

1. Tool call (MCP JSON-RPC)

POST https://plately.site/mcp
Authorization: Bearer <oauth-access-token>
Accept: application/json, text/event-stream
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "generate_diet_plan",
    "arguments": {
      "age": 34,
      "gender": "male",
      "weight_kg": 82,
      "height_cm": 176,
      "goal": "weight_loss",
      "activity_level": "moderate",
      "food_preference": "veg",
      "cuisine": "indian",
      "allergies": "peanuts",
      "diseases": "type 2 diabetes, fatty liver",
      "language": "en"
    }
  }
}

2. Tool response (result.structuredContent)

{
  "plan_id": "0a4c1e9b-77d2-4c1a-9c1f-8a2f6b0d1e42",
  "reused": false,
  "plan": {
    "title": "Diabetes-friendly veg day for steady weight loss",
    "daily_calories": 1750,
    "macros": { "protein_g": 110, "carbs_g": 170, "fat_g": 55 },
    "bmi": 26.5,
    "bmi_category": "Overweight",
    "tips": [
      "Eat protein first at every meal to blunt glucose spikes.",
      "Swap white rice for millet or quinoa at lunch.",
      "Walk 10–15 min after each main meal.",
      "Keep dinner light and finish by 8 PM.",
      "Hydrate: 8–10 glasses of water spread across the day."
    ],
    "health_highlights": [
      {
        "title": "Won't spike blood sugar",
        "detail": "Low-GI grains (oats, millet) + protein pairing keeps post-meal glucose steady.",
        "tag": "good",
        "related": "type 2 diabetes"
      },
      {
        "title": "Liver-friendly fats",
        "detail": "Uses olive oil, nuts (no peanuts), and avoids fried food to reduce fatty liver load.",
        "tag": "good",
        "related": "fatty liver"
      },
      {
        "title": "Peanut-free — safe",
        "detail": "All snacks and gravies use almonds/cashews or seeds instead of peanuts.",
        "tag": "safe",
        "related": "peanuts"
      },
      {
        "title": "Avoid added sugar & sweetened drinks",
        "detail": "Skip mithai, packaged juice, and sugary chai — they undo the calorie deficit.",
        "tag": "avoid",
        "related": "type 2 diabetes"
      }
    ],
    "days": [
      {
        "day": "Today",
        "meals": {
          "breakfast": {
            "name": "Vegetable oats chilla with mint chutney",
            "items": [
              "2 oats-besan chillas with grated carrot & spinach",
              "2 tbsp mint-coriander chutney (no peanuts)",
              "1 cup unsweetened masala chai"
            ],
            "calories": 380
          },
          "lunch": {
            "name": "Millet thali with dal & sabzi",
            "items": [
              "2 bajra rotis",
              "1 katori moong dal tadka",
              "1 katori bhindi masala (olive oil)",
              "1 bowl cucumber-tomato salad",
              "150 g low-fat curd"
            ],
            "calories": 560
          },
          "snack": {
            "name": "Roasted chana & almonds",
            "items": [
              "30 g roasted chana",
              "10 almonds",
              "1 cup green tea"
            ],
            "calories": 210
          },
          "dinner": {
            "name": "Paneer bhurji with sautéed greens",
            "items": [
              "120 g paneer bhurji (low-oil)",
              "1 bowl sautéed palak & beans",
              "1 multigrain roti",
              "Warm water with lemon"
            ],
            "calories": 600
          }
        }
      }
    ]
  }
}

Reused plans: if the user already generated a plan in the current 5 AM → 5 AM window, the same shape returns withreused: true and the existingplan_id — no new AI call is made.

Errors

• Not authenticated — the OAuth token is missing or expired; reconnect.
• You already have a plan for today... — generate_diet_plan returns your existing plan with reused: true.
• AI rate limit reached / AI credits exhausted — retry later or top up workspace credits.
• Validation errors follow Zod format: [{ path, message, code }].

Try it — live MCP call

Sign in required. This runs the same server logic the MCP tool executes and records the call in your profile's MCP history.